πŸ‡¨πŸ‡¦VancouverπŸ‡¨πŸ‡¦TorontoπŸ‡ΊπŸ‡ΈLos AngelesπŸ‡ΊπŸ‡ΈOrlandoπŸ‡ΊπŸ‡ΈMiami
1-855-KOO-TECH
KootechnikelKootechnikel
Insights Β· Field notes from the SOC
Plain-language briefings from the people watching the alerts.
Weekly Β· No spam
Back to News
Robotics & AutomationIndustry

Hugging Face and Pollen Robotics Launch Affordable Open-Source Humanoid Robot

AuthorZe Research Writer
Published
Read Time9 min read
Views0
Hugging Face and Pollen Robotics Launch Affordable Open-Source Humanoid Robot

Hugging Face and Pollen Robotics Launch Affordable Open-Source Humanoid Robot

Hugging Face partnered with French robotics company Pollen Robotics to release Reachy Mini, a $299 open-source humanoid robot designed to democratize robotics research and education.

## Executive Brief

Technical diagram showing vulnerability chain
Figure 1: Visual representation of the BeyondTrust vulnerability chain

Executive Brief

Hugging Face, the AI company known for its open-source machine learning platform, announced on June 1, 2025, a partnership with French robotics firm Pollen Robotics to release Reachy Mini, a compact humanoid robot priced at $299. The collaboration represents a significant push to make robotics hardware accessible to researchers, educators, and hobbyists who have traditionally faced barriers to entry due to high equipment costs.

Reachy Mini stands approximately 45 centimeters tall and features two articulated arms with gripper hands, a movable head with camera-based vision, and an open-source software stack built on the LeRobot framework. The robot connects to Hugging Face's ecosystem, allowing users to download pre-trained models, share their own trained behaviors, and collaborate on robotics research through the company's model hub.

The announcement positions Hugging Face as a player in physical AI systems, extending beyond its established role in language models and computer vision. Pollen Robotics, based in Bordeaux, France, brings manufacturing expertise from its larger Reachy robot platform, which has been deployed in research institutions and commercial settings since 2020.

Pre-orders opened on June 1, 2025, with shipping expected to begin in late 2025. The companies stated that all hardware designs, firmware, and software would be released under open-source licenses, enabling community modifications and third-party accessories.

What Happened

On June 1, 2025, Hugging Face and Pollen Robotics jointly announced the Reachy Mini humanoid robot through coordinated blog posts and press materials. The announcement came during a period of increased industry interest in embodied AI systems, where machine learning models control physical robots rather than operating purely in software.

Hugging Face CEO Clem Delangue described the project as an extension of the company's mission to democratize AI. According to the Hugging Face blog post, the partnership began in late 2024 when both companies identified a gap in the market for affordable, research-grade robotics hardware.

Pollen Robotics, founded in 2016, had previously developed the full-sized Reachy robot, which stands approximately 1.4 meters tall and costs several thousand dollars. The company's engineering team adapted the design for the smaller form factor while maintaining compatibility with the LeRobot software framework that Hugging Face released in early 2024.

The $299 price point places Reachy Mini significantly below comparable research robots. According to Ars Technica's coverage, similar platforms from academic robotics suppliers typically cost between $1,000 and $5,000, limiting their availability to well-funded research labs.

Authentication bypass flow diagram
Figure 2: How the authentication bypass vulnerability works

Key Claims and Evidence

Hugging Face and Pollen Robotics made several technical claims about Reachy Mini's capabilities:

Hardware specifications: The robot features 7 degrees of freedom per arm, a 2-axis neck for head movement, and integrated cameras in the head unit. The gripper hands can manipulate objects up to 200 grams, according to Pollen Robotics' technical documentation.

Software integration: Reachy Mini runs on the LeRobot framework, which Hugging Face released as an open-source project for robot learning. The framework supports imitation learning, where robots learn behaviors by observing human demonstrations, and reinforcement learning approaches.

Model sharing: Users can upload trained robot behaviors to the Hugging Face Hub, the same platform used for sharing language models and image generators. The companies claim this will accelerate research by allowing scientists to build on each other's work rather than starting from scratch.

Manufacturing timeline: Pollen Robotics stated that production would take place at their facility in France, with initial batch sizes of 1,000 units. The company has experience manufacturing the larger Reachy robot and claims the supply chain is established.

Pros and Opportunities

The Reachy Mini release addresses several longstanding challenges in robotics research and education:

Cost accessibility: At $299, the robot falls within the budget range of individual researchers, small labs, and educational institutions. Graduate students and hobbyists who previously could not afford research-grade hardware gain access to a capable platform.

Open-source ecosystem: The combination of open hardware designs and the LeRobot software framework creates opportunities for community-driven development. Researchers can modify the robot's physical design, contribute software improvements, and share trained models.

Educational applications: Universities and coding bootcamps can incorporate physical robotics into curricula without major equipment investments. The robot's size makes it suitable for classroom environments where space is limited.

Research reproducibility: When multiple labs use the same hardware platform, research results become more comparable. The robotics field has historically struggled with reproducibility because each lab often builds custom hardware.

Integration with existing tools: Hugging Face's platform already hosts thousands of machine learning models. Reachy Mini users can potentially adapt vision models, language models, and other AI components for robotics applications.

Privilege escalation process
Figure 3: Privilege escalation from user to SYSTEM level

Cons, Risks, and Limitations

Several factors constrain Reachy Mini's capabilities and market position:

Payload limitations: The 200-gram manipulation limit restricts the robot to lightweight objects. Many practical robotics applications require handling heavier items, limiting Reachy Mini to research and educational contexts rather than commercial deployment.

Scale constraints: The 45-centimeter height means the robot operates in a limited workspace. Tasks requiring human-scale reach or interaction with standard furniture and fixtures fall outside its capabilities.

Unproven at scale: While Pollen Robotics has manufactured the larger Reachy robot, producing thousands of units at the $299 price point represents a different manufacturing challenge. Supply chain issues or quality control problems could affect early adopters.

Software maturity: The LeRobot framework, while open-source, is relatively new. Documentation, tutorials, and community support are still developing. Users without robotics experience may face a steep learning curve.

Competition from simulation: Some researchers argue that simulated environments provide faster iteration cycles than physical robots. The value proposition of affordable hardware depends on tasks that require real-world physics and sensor data.

How the Technology Works

Reachy Mini combines mechanical design, embedded computing, and machine learning software into an integrated system:

Mechanical structure: The robot's skeleton uses 3D-printed components and off-the-shelf servo motors. Each arm contains seven servo motors providing rotation at the shoulder, elbow, and wrist joints. The gripper mechanism uses a single motor to open and close the fingers.

Sensing systems: Two cameras mounted in the head provide stereo vision for depth perception. The robot includes inertial measurement units for balance sensing and position encoders on each joint for proprioception. These sensors feed data to the onboard computer for processing.

Computing architecture: A Raspberry Pi-class single-board computer handles low-level motor control and sensor processing. For computationally intensive tasks like neural network inference, the robot can offload processing to a connected laptop or cloud server via WiFi.

LeRobot framework: The software stack abstracts hardware details behind a Python API. Researchers can write control code without managing low-level motor commands. The framework includes tools for recording human demonstrations, training imitation learning models, and deploying trained policies to the robot.

Model deployment: Trained models stored on the Hugging Face Hub can be downloaded and executed on Reachy Mini with minimal configuration. The standardized model format means behaviors developed by one researcher can run on another researcher's robot without modification.

Technical context for expert readers: The LeRobot framework uses PyTorch for neural network training and supports common architectures including transformers and diffusion policies. The robot's joint space is represented as a 16-dimensional vector (7 joints per arm plus 2 neck joints), with actions specified as target joint angles or velocities.

Broader Industry Implications

The Reachy Mini announcement reflects several trends in the robotics and AI industries:

Commoditization of robotics hardware: As manufacturing techniques improve and component costs decrease, research-grade robots are becoming accessible to a wider audience. The $299 price point, while not consumer-level, represents a significant reduction from historical norms.

Platform strategies in AI: Hugging Face's expansion into hardware follows a pattern seen in other technology sectors. By providing both software tools and compatible hardware, the company creates a more integrated ecosystem that increases user engagement with its platform.

Open-source hardware movement: The decision to release hardware designs under open licenses aligns with broader trends in scientific instrumentation. Open-source lab equipment, 3D printers, and electronics have gained traction in research communities seeking to reduce costs and increase customization.

Embodied AI research: The machine learning community has shown increased interest in robots that interact with physical environments. Language models and image generators have achieved impressive results, but transferring those capabilities to physical systems remains an active research area.

Geographic distribution of robotics: Pollen Robotics' location in France highlights that robotics innovation occurs globally, not solely in traditional hubs like Silicon Valley or Shenzhen. European robotics companies have carved out niches in research platforms and collaborative robots.

Confirmed Facts vs. Open Questions

Confirmed:

  • Reachy Mini is priced at $299 for pre-orders opening June 1, 2025
  • The robot features two 7-DOF arms and a 2-axis head
  • Hardware designs and software will be released under open-source licenses
  • Pollen Robotics will manufacture the robots in France
  • The robot integrates with the LeRobot framework and Hugging Face Hub

Unconfirmed or unclear:

  • Exact shipping dates for pre-orders (stated as "late 2025")
  • Total production capacity and whether demand will exceed supply
  • Long-term support and warranty terms
  • Whether the $299 price is sustainable or represents an introductory offer
  • Performance benchmarks compared to more expensive research platforms

What to Watch Next

Several developments will indicate whether Reachy Mini achieves its stated goals:

Pre-order volume: The number of units ordered in the initial period will signal market demand for affordable research robots. High demand could validate the business model but also strain manufacturing capacity.

Community contributions: Activity on the LeRobot GitHub repository and Hugging Face Hub will show whether researchers adopt the platform. Model uploads, bug fixes, and documentation improvements indicate healthy community engagement.

Academic adoption: Announcements from universities incorporating Reachy Mini into courses or research programs would demonstrate institutional acceptance. Conference papers using the platform would further validate its research utility.

Competitor responses: Other robotics companies may adjust pricing or release competing products. The market's reaction will clarify whether $299 represents a new baseline for research robot pricing.

Manufacturing execution: Pollen Robotics' ability to deliver units on schedule and at the stated quality level will determine early adopter satisfaction. Delays or quality issues could undermine the project's momentum.

Sources

  1. Ars Technica - "Hugging Face launches $299 open-source humanoid robot" - June 1, 2025 - https://arstechnica.com/ai/2025/06/hugging-face-launches-299-open-source-humanoid-robot/

  2. Hugging Face Blog - "Introducing Reachy Mini" - June 1, 2025 - https://huggingface.co/blog/reachy-mini

  3. Pollen Robotics - "Reachy Mini Product Page" - June 1, 2025 - https://www.pollen-robotics.com/reachy-mini/

Sources & References

Related Topics

roboticsopen-sourcehugging-facehumanoidai